If we go with the above, we need to think of how to name things…see https://www.pantsbuild.org/v2.0/docs/rules-api-testing for the four approaches I think we should document: 1. normal Python tests 2.

run_rule()

/ aka unit tests for rules 3.

RuleRunner()

/ aka integration tests for rules 4.

run_pants()

/ aka integration tests for Pants (end-to-end tests?)

run_rule()

and

RuleRunner()

sound way too similar. One possibility is to consolidate #2 and #3 a bit so that you always use

RuleRunner

for them. For #2, you use a classmethod like

RuleRunner.run_rule_with_mocks()

, whereas with #3 you run an instance method

rule_runner.request_product()